EVOLUTION INTERNATIONAL JOURNAL OF ORGANIC EVOLUTION PUBLISHED BY THE SOCIETY FOR THE STUDY OF EVOLUTION

Vol. 55 July 2001 No. 7

Evolution, 55(7), 2001, pp. 1269±1282

PERSPECTIVE: EVOLUTION OF FLOWER COLOR IN THE DESERT ANNUAL PARRYAE: WRIGHT REVISITED

DOUGLAS W. SCHEMSKE1,2 AND PAULETTE BIERZYCHUDEK3 1Department of , Box 355325, University of Washington, Seattle, Washington 98195-5325 E-mail: [email protected] 3Department of Biology, Box 53, Lewis and Clark College, Portland, Oregon 97219-7899 E-mail: [email protected]

Abstract. Linanthus parryae, a diminutive desert annual with white or blue ¯owers, has been the focus of a long- standing debate among evolutionary biologists. At issue is whether the ¯ower color in this species is the product of random , as Sewall Wright argued, or of , as proposed by Carl Epling and his colleagues. Our long-term studies of three polymorphic populations in the Mojave Desert demonstrate that ¯ower color is subject to selection that varies in both time and space in its direction and magnitude. For all sites taken together, blue-¯owered produced more seeds than white-¯owered plants in years of relatively low seed pro- duction, whereas white-¯owered plants had higher ®tness in years of high seed production. Evidence of selection on ¯ower color was found in two of the three study sites. Differences in ®tness between the color morphs were sometimes large, with selection coef®cients as high as 0.60 in some years. Our longest period of observations was at Pearblossom site 1, where plants reached appreciable densities in seven of the 11 years of study. Here we found signi®cant differences in the seed production of the color morphs in six years, with three years of blue advantage and three years of white advantage. For all sites taken together, total spring precipitation (March and April) was positively correlated with both absolute and relative seed production of the color morphs. At Pearblossom site 1, blue-¯owered plants typically had a ®tness advantage in years of low spring precipitation, whereas white-¯owered plants had a ®tness advantage in years of high spring precipitation. This temporal variation in selection may contribute to the maintenance of the ¯ower-color polymorphism at Pearblossom site 1, whereas gene ¯ow from neighboring populations is proposed as the principal factor maintaining the polymorphism at the other study sites. We found no signi®cant differences between the color morphs in pollinator visitation rate or in their carbon isotope ratio, a measure of water-use ef®ciency. Although the mechanism of selection remains elusive, our results refute Wright's conclusion that the ¯ower color polymorphism in L. parryae is an example of isolation by distance, a key component of his shifting balance theory of evolution.

Key words. Flower color, genetic drift, isolation by distance, natural selection, polymorphism, shifting balance theory, Sewall Wright.

Received August 9, 2000. Accepted April 4, 2001.

subject of debate for decades (Endler 1986; Provine 1986; Coyne et al. 1997). Many of the early architects of the Modern Linanthus had great signi®cance for Wright as his ®rst Synthesis maintained that adaptive evolution occurred prin- and best example of isolation by distance, a theory that cipally by natural selection within populations (Fisher 1930; Wright considered to be one of his most important con- tributions to evolutionary biology. Haldane 1932; Mayr 1942; Fisher and Ford 1947). This view Provine (1986, p. 485) was perhaps best exempli®ed by Fisher (1922), who believed that adaptation resulted from the ®xation of many individ- ually favorable mutations, each with small phenotypic ef- The relative contribution of natural selection and random fects. In contrast, Sewall Wright suggested that adaptation genetic drift to the evolution of adaptations has been the may often involve the stochastic process of genetic drift in addition to natural selection. 2 Present address: Department of Botany and Pathology, In Wright's shifting balance theory of evolution (Wright Michigan State University, East Lansing, Michigan 48824-1312. 1931, 1978, 1982), populations acquire favorable gene com- 1269 ᭧ 2001 The Society for the Study of Evolution. All rights reserved. 1270 D. W. SCHEMSKE AND P. BIERZYCHUDEK binations by a three-phase process involving both genetic over the importance of genetic drift and natural selection in drift and selection. In phase 1, genetic drift within demes L. parryae and by the pivotal role this species has played in decreases mean ®tness, allowing them to descend adaptive the development of evolutionary theory (Provine 1986). We peaks and to move across adaptive valleys. In phase 2, mass ®nd that natural selection on ¯ower color in L. parryae can selection elevates demes to new, higher adaptive peaks; and be strong, but that the direction and intensity of selection in phase 3, interdeme selection allows demes with higher vary in both time and space. Our studies suggest that the ®tness to spread across the adaptive landscape. Phase 3 of frequencies of blue- and white-¯owered L. parryae are largely the shifting balance has received the most attention (Crow a product of temporal and spatial heterogeneity in local se- et al. 1990; Wade and Goodnight 1991; Barton 1992; Kon- lection pressuresÐnot of isolation by distance, as concluded drashov 1992), but recent theoretical work has focused on by Wright (Wright 1943a, 1978). phase 1, and the potential role of drift in the formation of We ®rst describe the biology of L. parryae, then provide adaptive gene complexes (Goodnight 1995; Coyne et al. a chronological account of the nearly 40 years of research 1997). conducted on this species. This background allows us to place Wright (1931) suggested that phase 1 of the shifting bal- our approaches and interpretations in a historical context. We ance would occur most effectively in species subdivided into then outline the research objectives of our study. small, partially isolated demes. He proposed that restricted gene ¯ow in continuously distributed populations was the Natural History of Linanthus parryae principal means of attaining such population structure. Linanthus parryae () is a diminutive (Ͻ 3 Wright found support for this theory, which he termed ``iso- cm tall) annual native to the southern and western edges of lation by distance'' (Wright 1943a), from ®eld studies of a the Mojave Desert in California (Patterson 1993). It has large ¯ower color polymorphism in the annual plant Linanthus par- ¯owers (1.0±1.5 cm long, 2 cm wide) and a few tiny, nee- ryae. These studies were initiated by Theodosius Dobzhansky dlelike leaves, with calyces comprising most of the photo- and Carl Epling in 1941, and were continued by Epling and synthetic area. Epling and Dobzhansky (1942) described two his colleagues over a period of more than 20 years (Epling ¯ower morphs, blue and white, with the white form most and Dobzhansky 1942; Epling et al. 1960). According to common throughout the range. Crossing studies carried out Wright, this was ``probably the most thoroughly studied case by Epling and his colleagues provided evidence that ¯ower of a simple Mendelian polymorphism, within and among pop- color is controlled by a single gene, with blue dominant to ulations'' (Wright 1978, p. 194). white (Epling et al. 1960; Wright 1978), although the details Reporting the results from a single year of ®eld observa- of these experiments were never published (H. Lewis, pers. tions, Epling and Dobzhansky (1942) at ®rst concluded that comm.). The blue ¯owers in some populations of L. parryae the spatial pattern of ¯ower color in L. parryae was best display considerable variation in intensity (e.g., Antelope explained by random genetic drift, and this was supported Valley, see below), suggesting the presence of modi®er by Wright's analysis of their data published soon thereafter genes. (Wright 1943b). However, following many years of detailed Seed germination occurs after winter rains, plants ¯ower ®eld observations and experiments, Epling et al. (1960) re- in early to late April, and seeds mature in late May to early jected this earlier conclusion, and instead proposed that nat- June. The seeds are passively dispersed, falling from the cap- ural selection, not genetic drift, maintained the polymor- sule as it dries. There is no germination in years with low phism. Wright again reanalyzed the data collected by Epling rainfall (Epling et al. 1960; pers. obs.), and ungerminated and his colleagues, and disagreed with their new interpre- seeds remain dormant in the soil. In wet years plants can be tation, concluding once more that genetic drift was of primary exceptionally dense, hence the common name ``desert importance (Wright 1978). snow.'' Flower and seed number per plant are highly variable, This wide disparity in opinions is re¯ected in the varied with plants in dry years producing an average of one or two interpretations of the L. parryae studies presented over the ¯owers and 10±30 seeds, whereas in good years reproductive years by other workers. Stebbins (1950) cited L. parryae as output may increase by nearly 10-fold. The sole pollinator an example of isolation by distance. In contrast, Mayr (1965) is the beetle Trichochorous sp. (Melyridae), and the ¯owers concluded that the evidence collected by Epling et al. (1960) are self-incompatible (Epling et al. 1960; D. W. Schemske showed that the ¯ower color polymorphism in L. parryae was and P. Bierzychudek, pers. obs.). maintained by selection. Grant (1981, p. 25) concluded that the spatial distribution of the color morphs ``may well have Background to the Debate developed as a result of restricted gene ¯ow and neighbor- hood size, and without any role of environmental selection, The ®rst research carried out on L. parryae was a 1941 although this is by no means certain.'' Waser and Price (1985) census of ¯owering plants by Epling and Dobzhansky (1942). questioned Wright's conclusion that spatial patterns such as They recorded the numbers of blue- and white-¯owered those exempli®ed by L. parryae are due to isolation by dis- plants at four different stations distributed every 0.5 mi along tance and suggested the alternative hypothesis of local se- a 200-mi grid of roads traversing the Mojave Desert. The lection. Finally, Levin (1988, p. 314) suggested that ``one white morph was more abundant overall; 78% of their sam- highly probable case for local random differentiation is de- ples consisted only of white-¯owered individuals (data from scribed by the spatial distribution of ¯ower polymorphism table 1 of Epling and Dobzhansky 1942). For this reason, in the outcrossing annual Linanthus parryae.'' they designated certain geographical regions as ``variable Our research was motivated by this long-standing debate areas'' and focused the rest of their analysis on these sam- LINANTHUS FLOWER COLOR 1271 pling stations. Most of the samples in these variable areas ever, that they saw no evidence that such selection was ac- contained at least some blue plants, and about 10% were tually taking place. They concluded that the distribution of monomorphic for blue. Based on the data for the variable blue- and white-¯owered plants was due to ``selection op- areas, they found that the frequency distribution of the per- erating at relatively low intensity or intermittently'' (Epling centage blue plants was U-shaped, and concluded that this et al. 1960, p. 254). ``resembles Wright's curves for the distribution of gene fre- In 1978 Wright returned to the L. parryae problem, sum- quencies in effectively small populations'' (Epling and Dob- marizing earlier results and adding some new analyses zhansky 1942, p. 331). This was consistent with Wright's (Wright 1978). He reanalyzed the census data in Epling et (1931) expectation that in the absence of selection, mutation, al. (1960) and subsequent unpublished data collected by and migration, alternate alleles in different populations will Epling and his colleagues. Wright created 52 groups, each become ®xed by random genetic drift. comprised of ®ve quadrats, and conducted ␹2-tests of the Wright published his own analysis of Epling and Dob- change in morph frequency across ®ve different time periods zhansky's data (Wright 1943b). He calculated F, ``the amount spanning 22 years. He showed that there were few signi®cant of differentiation among groups taken at random from the changes, and that these were ``widely scattered,'' leading him whole'' (Wright 1943b, p. 123), then estimated the effective to the conclusion that ``In the main, the changes were clearly population size that would be required to produce the ob- of the nature of random genetic drift'' (Wright 1978, p. 222). served F. Wright's classic paper ``Isolation by Distance'' Wright again calculated F-statistics from the data collected (Wright 1943a) provided the theoretical development re- by Epling and Dobzhansky (1942), and plotted FIS and FST quired for this analysis, and was published in tandem with as a function of the size of the region sampled. He concluded his L. parryae paper (Wright 1943b). For the case of complete that ``these curves are fairly typical of those expected where outcrossing, with the blue allele dominant to the white, diversi®cation at all levels is built up merely by isolation by Wright estimated an effective population size on the order distance'' (Wright 1978, p. 203). Wright conceded that ge- of 14±25 individuals. He also calculated F for a hierarchy of netic drift could not explain the unexpectedly high degree of spatial subdivisions, ranging from square feet to hundreds of differentiation at the largest spatial scales: ``It is barely pos- square miles, and plotted F as a function of the size of the sible that the profound differentiation of the four primary subdivision. Wright concluded that ``The distribution of blue subdivisions is built up from random drift of neighborhoods. and white can be accounted for most easily by supposing This area, containing millions of neighborhoods, is probably that most of the differentiation of the smaller categories [sub- so great, however, that such a buildup would almost certainly divisions] is random in character'' (Wright 1943b, p. 155). be prevented by recurrent mutation'' (Wright 1978, p. 223). Carl Epling maintained his interest in L. parryae, and in Instead, he hypothesized that white-¯owered plants had ``a collaboration with and Francis Ball, initiated slight, general selective advantage,'' with selection coef®- a long-term study of L. parryae in 1944 that resulted in the cients in the range of s ϭ 0.01±0.0001, and that this was third publication on the biology of this species (Epling et al. more likely due to ``signi®cant environmental differences 1960). Their work provided two major ®ndings that chal- among broad regions than among the many thousands of lenged Wright's conclusions. First, they demonstrated that small areas that differ apparently at random in blue frequen- L. parryae seeds could remain viable in the soil for at least cy'' (Wright 1978, p. 223). seven years, and probably much longer, thereby producing a In summary, Wright believed that random genetic drift large ``effective breeding group'' (Epling et al. 1960, p. 254). prevailed at the level of demes, and at all but the largest This result was based on an extraordinary ®eld experiment spatial scales, where selection played a modest role. This is in which ¯owering plants were counted, then removed from at odds with Epling et al. (1960), who concluded that even marked plots for seven successive years. Over this time pe- small-scale differences in morph frequency were due to nat- riod more than 16,000 plants were observed, all of which had ural selection, not to genetic drift. emerged from the seed bank. Second, by establishing per- manent census transects, they demonstrated that morph fre- Research Objectives quencies were far more stable than was expected from Wright's estimate of the effective population size. Their tran- We have conducted long-term ®eld studies to investigate sect was 10 ft wide and 0.5 m long, divided into 260, 10 ft two questions concerning ¯ower color evolution in L. par- ϫ 10 ft quadrats. In each quadrat they recorded the number ryae: (1) Is the polymorphism observed within populations of blue- and white-¯owered plants in the single square-foot selectively neutral? (2) Are the differences in ¯ower color sample with the highest plant density. They monitored this observed between populations due to spatially varying natural transect every year from 1944 to 1958, except for years when selection? Here we report on the evidence for selection in L. parryae was absent (two years) or rare (three years). They polymorphic populations. Our investigations of ¯ower color found considerable spatial variation in morph frequencies, evolution in nearly monomorphic populations will be pre- but very little temporal variation, and concluded that ``if sented in a later paper. genetic drift has played a role, it has been of only local Whereas both Wright and Epling et al. (1960) relied on consequence and not persistent in its effects'' (Epling et al. indirect approaches to reach their conclusions, we estimated 1960, p. 254). Instead, they hypothesized ``an intense local selection on ¯ower color directly, by measuring seed pro- selection because the blues are concentrated in certain areas duction of the two morphs. We made this comparison over and because persisting clines of blue and white frequencies a series of 11 years, because the maintenance of a polymor- have been found'' (p. 254). Epling et al. acknowledged, how- phism by directional selection requires, at the very least, that 1272 D. W. SCHEMSKE AND P. BIERZYCHUDEK the direction of selection ¯uctuate over time (Turelli et al. south of the intersection of Lancaster Road and 200th Street 2001). Thus, we determined whether the relative ®tness of West, west of the town of Lancaster, California, and 55 km the two morphs varies from one year to another, and if so, WNW of the Pearblossom locality. This small, mixed pop- what mechanisms might be responsible for any such varia- ulation of L. parryae is not more than 500 m in diameter and tion. is con®ned to the base of a narrow valley; the nearest mono- In addition to measuring ®tnesses of blue- and white-¯ow- morphic L. parryae population is at least 0.5 km away. ered morphs of L. parryae and how these values may change For most of the data we have collected, we provide three over time, we also investigated some possible agents of se- levels of analysis, at increasingly more restricted spatial lection. We ®rst asked if Linanthus's pollinators might pref- scales. First, we combine all three populations and all avail- erentially visit one ¯ower color over another, and if so, able years to provide the largest possible sample and most whether the preferred morph changes over time. Evidence general result; second, we present the combined data from for pollinator preference in polymorphic species exists from the two Pearblossom populations, a smaller and more local- other studies (Kay 1978; Brown and Clegg 1984; Stanton ized sample; and third, we provide an analysis for Pearblos- 1987; Stanton et al. 1989). som 1 only, the site that we have studied for the longest time We also asked if ¯ower color might have indirect effects period and the one that overlaps the area studied by Epling on plant ®tness. In several other California Linanthus spp. and his colleagues. that produce both white and pigmented ¯owers, the pig- mented morph is typically associated with stressful environ- Morph Frequency, Density, and Reproduction ments, for example serpentine soils or habitats with a rapid decline in soil moisture through the growing season (D. W. To measure morph frequency and density and to measure Schemske, pers. obs.). Pleiotropic effects of ¯ower color are ®tness of the two morphs, at each site we established three known from other species (Burdon et al. 1983; Schoen et al. 50m ϫ 1 m permanent transects placed roughly parallel to 1984; Ernst 1987; Taylor and Jorgensen 1992; Rausher and one another and approximately 7 m apart. In each year of Fry 1993). Of particular interest is the ®nding by Levin and the study, during the peak of ¯owering (mid±late April), we Brack (1995) that white-¯owered individuals of Phlox drum- censused the numbers of ¯owering individuals of each ¯ower 2 mondii, a species that is in the same family (Polemoniaceae) color within each 1-m quadrat of every transect, 150 m at as Linanthus, had lower ®tness than pink-¯owered morphs each site. In years when plants were present, we placed in- and that this ®tness difference was due to pleiotropic effects. dividually numbered plastic stakes next to a subset of indi- In a desert environment, the ability to tolerate soil moisture viduals from within these transects and collected their seeds. de®cits is of obvious survival value. Thus, we asked whether We marked plants in pairs, one blue and one white, that were plants with different colored ¯owers may also differ in their nearest neighbors. Blue plants were always less frequent than water use ef®ciency, such that soil moisture levels are ex- white, so we typically marked every blue plant in each quad- perienced differently by the two morphs. To test this pos- rat as well as that plant's nearest white neighbor. In years of sibility, we analyzed carbon isotope ratios of dried plant tis- low plant density, we occasionally marked plants just outside sue to infer the water use ef®ciency of the two morphs (Eh- the quadrats to increase the sample size, but the distance leringer 1988). between paired plants was never greater than 1 m and typ- ically was on the order of a few centimeters. In years when MATERIALS AND METHODS plant density was particularly high and the number of blue plants per subquadrat (0.25 m2) was greater than ®ve, we Study Sites marked only ®ve randomly selected pairs per subquadrat. We Our work was conducted in two different localities north marked an average of 340 (range 88±521) plant pairs at a of the San Gabriel Mountains in Los Angeles County, Cal- site each year. ifornia. The two Pearblossom sites, which are about 1.5 km We collected each marked plant once its seeds had matured apart, are both located along Avenue V, 5 km west of Pear- (late May, early June) into an individual paper envelope. Each blossom (elevation 900 m), between the Pearblossom High- plant was examined under a dissecting microscope in the way (CA 138) and 96th Street East; this is about midway laboratory, where we counted the number of ¯owers and between the towns of Pearblossom and Littlerock. This cre- seeds produced. We then returned the seeds from each plant osote bush desert has L. parryae populations that differ wide- to the exact position where they were collected, still marked ly in the frequency of the two color morphs, including nearly by its plastic stake. Seeds were returned as soon as possible monomorphic populations as well as the two polymorphic after counting, usually in July of that year. In total, we count- ones that we report on here. Pearblossom 1, the site further ed 41,718 plants in our census transects and 713,364 seeds to the east, overlaps with the study sites used by Carl Epling on marked plants. and his colleagues in their studies (H. Lewis, pers. comm.; We used the seed number of ¯owering plants as our mea- Epling and Dobzhansky 1942; Epling et al. 1960). This site sure of ®tness in this outcrossing, hermaphroditic annual. Our maintains its variable nature for at least 0.5 km in all direc- failure to include the male component of ®tness will doubtless tions. In contrast, Pearblossom 2 is closely bounded by nearly cause some error in our estimates of the intensity of selection monomorphic populations, with a mostly blue-¯owered pop- on ¯ower color. We also compared ¯ower number of the two ulation 50 m from its western edge and a mostly white-¯ow- morphs to attempt to understand the causes of any differences ered population 50 m from its eastern edge. in seed production between them. The Antelope Valley locality (elevation 900 m) is 3 km Although it is clearly possible that selection may operate LINANTHUS FLOWER COLOR 1273 prior to ¯owering, our ®eld observations support the ®ndings to our Pearblossom sites than the Palmdale station, its op- of Epling et al. (1960) that most seedlings survive to ¯ower. eration has been inconsistent, resulting in missing data for Because all withered plants remain standing, dry and brittle, some months of the study. The precipitation recorded during for many months, evidence of mortality was not easy to over- the period of study at Palmdale and Pearblossom was highly look. Furthermore, the seeds from monomorphic populations correlated for the entire growing season (January±March, r of each morph that we germinated in the greenhouse showed ϭ 0.95, n ϭ 9, P Ͻ 0.0001) and for each individual month no obvious differences in their germination rates (P. Bier- (January, r ϭ 0.96, n ϭ 11, P Ͻ 0.0001; February, r ϭ 0.99, zychudek and D. W. Schemske, pers. obs.). We attempted to n ϭ 10, P Ͻ 0.0001; March, r ϭ 0.90, n ϭ 10, P Ͻ 0.0001; compare the longevity of seeds of the two morphs in the seed April, r ϭ 0.83, n ϭ 10, P Ͻ 0.01). bank, but this experiment was vandalized and yielded no Finally, desert precipitation is notoriously variable, so the useful information. 11 years of our study may not be representative of longer- We compared ¯ower number and seed number of blue- and term trends in precipitation. We compared the rainfall pattern white-¯owered plants by Wilcoxon paired two-sample tests. observed during our study to that for the period from 1941 Sequential Bonferroni tests (Rice 1989) were carried out sep- to 1987, that is, from the ®rst year of Epling and Dob- arately for each site, with a tablewide ␣ level of 0.05. We zhansky's original study until the year before our study. Com- provide the statistical results for individual tests of site-year parisons were made for those monthly combinations of rain- combinations both with and without the Bonferroni correc- fall found to best explain the observed variation in plant tion. Flower number could not be determined for plants that density and plant performance. had experienced severe herbivory, so these plants were omit- ted from the analyses of ¯ower number, but not from the Possible Mechanisms of Selection seed number analysis. Plants that had ¯owered but produced no seeds, due either to herbivory or to fruiting failure, were Pollinator preferences included in the analyses of seed number. The seeds usually To determine whether pollinators preferentially visited fell from the fruits once the plants were placed in the col- blue- or white-¯owered L. parryae, observers watched a ran- lecting envelope, so it was not possible to estimate the seed domly chosen plant for 30 sec, counting the number of beetle production of individual fruits within a plant. This prevented visits that took place during that time. At the end of the us from estimating the contributions of ¯ower number per period, the observer moved to the nearest plant of the other plant and of seed number per fruit to the variation observed color for an additional 30 sec. This sequence of alternating- in total seed number per plant. Sample sizes varied among color observations continued throughout the period that pol- years and sites, due primarily to temporal variation in plant linator activity was most intense, between 1000 and 1500 h. density. Beetles did not appear to be disturbed by the presence of We used Fisher's combined probability test (Fisher 1954) observers. It was usually impossible to follow beetles as they to assess the overall signi®cance of the statistical analyses moved between plants, so we could not estimate the degree conducted on ¯ower and seed number. This method is used of ¯ower-color preference of individual pollinators. All ob- when the combined data are unlikely to meet the assumptions servations were made near the Pearblossom 2 site, because of more complex, parametric analyses such as two-way AN- pollinator densities were consistently higher there than at the OVA (Sokal and Rohlf 1994). For each site, we calculated other sites. Pollinator observations were carried out during Ϫ2 ⌺ ln Pk , where Pk is the probability level of the kth the peak of the blooming season, for 33.7 observer-hours in independent test of signi®cance. The combined probability 1991 and 74.0 observer-hours in 1992. Because we observed is distributed as ␹2 with 2k degrees of freedom (Sokal and no evidence of heterogeneity in the visitation pattern based Rohlf 1994). on date or time of day, we pooled the observations from each year. We used a G-test with one degree of freedom to de- Precipitation and Plant Performance termine if the number of visits observed was different from the null expectation of equal visitation to the two morphs. To determine whether there was an association between plant performance and precipitation levels, we identi®ed the times of year during which precipitation was most strongly Water-use ef®ciency correlated with plant density and seed production. We did A number of studies have demonstrated that the relative this by calculating the Spearman rank correlation coef®cient ratio of 13C/12C, typically expressed as ␦13C, provides an between these variables and precipitation for each month of indirect measure of plant water use ef®ciency (WUE; Ehler- the growing season (January±April) separately, as well as for inger et al. 1987; Farquhar et al. 1988; Martin and Thor- successive combinations of months (e.g., February ϩ March, stenson 1988). Plants with high WUE possess higher (more February ϩ March ϩ April, etc.). We used data for the period positive) carbon isotope ratios, a relationship that results from 1988±1998, but for the analysis of seed production, we ex- the strong discrimination by the initial carboxylating enzyme cluded years in which ¯owering plants were absent or nearly ribulose bisphosphate carboxylase against heavier carbon iso- so (1990, 1994, 1996, and 1997). topes (Ehleringer 1988). To determine if ¯ower color in L. We obtained precipitation data from the Palmdale, Cali- parryae has a pleiotropic effect on WUE, we compared the fornia, recording station, located 18 km NW of the Pear- carbon isotope ratios (␦13C) for blue- and white-¯owered blossom locality and 43 km ESE of the Antelope Valley plants collected from Pearblossom 1 in 1998. The expense locality. Although the Pearblossom recording station is closer of these analyses required that we create pooled samples of 1274 D. W. SCHEMSKE AND P. BIERZYCHUDEK

TABLE 1. Density and frequency of blue- and white-¯owered Linanthus parryae at the three study sites in the Mojave Desert. Standard deviation for percentage blue calculated according to Snedecor and Cochran (1967).

Number of plants1 Density Site Year Blue White Total (plants/m2) % blue (SD) Pearblossom 1 1988 371 2383 2754 18.4 13.5 (0.65) 1989 32 201 233 1.6 13.7 (2.25) 1990 0 0 0 0 1991 138 732 870 5.8 15.9 (1.24) 1992 479 2789 3268 21.8 14.7 (0.62) 1993 200 1726 1926 12.8 10.4 (0.70) 1994 2 5 7 0.047 28.6 (17.08) 1995 1153 8818 9971 66.5 11.6 (0.32) 1996 ϳ0 ϳ0 ϳ0 ϳ0 1997 0 0 0 0 1998 308 3124 3432 22.9 9.0 (0.49) Pearblossom 2 1991 596 902 1498 10.0 39.8 (1.26) 1992 603 1074 1677 11.2 36.0 (1.17) 1993 184 424 608 4.1 30.3 (1.86) 1994 0 0 0 0 Antelope Valley 1990 336 1244 1580 10.5 21.3 (1.03) 1991 520 2735 3255 21.7 16.0 (0.64) 1992 490 4925 5415 36.1 9.0 (0.39) 1993 1185 4039 5224 34.8 22.7 (0.56) 1 Pooled date from three, 1 m ϫ 50 m transects at each site. individuals having the same ¯ower color. We paired a pooled Antelope Valley, morph frequency ¯uctuated somewhat sample of 10 blue individuals with another pooled sample of more; here the frequency of the blue morph varied from 9.0 10 white individuals. Each of the 10 white individuals was Ϯ 0.4% in 1992 to 22.7 Ϯ 0.6% in 1993 (Table 1). At both the nearest neighbor of one of the 10 blue individuals in its Pearblossom sites the frequency of blue-¯owered plants de- paired sample. The 10 pairs of blue- and white-¯owered in- clined steadily from 1991 to 1993. dividuals were collected from within a few meters of one another within one of the transects. This process was repeated Plant Fitness six times from each of the three transects, for a total of 18 pairs of pooled samples. By pooling tissue from 10 different Mean ¯ower production for all sites and years is given in plants to construct each sample, we reduced the variation Table 2. Individual tests of signi®cance revealed that blue- among samples and increased the precision in estimating the ¯owered plants produced signi®cantly more ¯owers than mean. Each sample was prepared for analysis by grinding the whites at Pearblossom 1 in 1988 and at Antelope Valley in roots, stems, and leaves of individual plants to a ®ne powder. 1990, whereas white-¯owered plants produced signi®cantly Flowers and fruits were not included. The carbon isotope more ¯owers than blues at Pearblossom 1 in 1998 and at ratio of plant tissue was determined by the Stable Isotope Pearblossom 2 in 1991. All but the 1990 comparison at An- Ratio Facility for Environmental Research at the University telope Valley remain signi®cant after the Bonferroni correc- of Utah (http://ecophys.biology.utah.edu/sirfer). We used tion (Table 2). There was considerable temporal variation in paired t-tests to compare the carbon isotope ratios of the two ¯ower number at all sites, although the range of mean values morphs. at Pearblossom (1.2±15.0 ¯owers/plant) was much greater than that at Antelope Valley (1.5±4.9 ¯owers/plant). Fisher's RESULTS combined probability test revealed a signi®cant overall effect of ¯ower color on ¯ower number at Pearblossom 1 (␹2 ϭ Plant Density and Morph Frequency 41.2, df ϭ 14, P Ͻ 0.0001), but not at Pearblossom 2 (␹2 ϭ Each of our three study sites displayed striking temporal 11.0, df ϭ 6, P ϭ 0.090) or at Antelope Valley (␹2 ϭ 14.6, variation in plant density (Table 1). For example, at Pear- df ϭ 8, P ϭ 0.071). blossom 1, mean annual plant density ranged from zero to Individual tests of signi®cance revealed that the seed pro- Ͼ 60 plants/m2 over the 11 years of observation. Mean den- duction of the two color morphs was signi®cantly different sities at Antelope Valley varied less from year to year than in six of the seven years that plants ¯owered at Pearblossom at the Pearblossom sites, and moderate ¯owering plant den- 1 (Table 3; Fig. 1), and in one of three years at Pearblossom sities were found at this site in 1990, a year when Pearblos- 2 (Table 3; Fig. 2A). At Pearblossom 1, blue-¯owered plants som 1 did not produce any plants. produced signi®cantly more seeds than whites in 1988, 1992, The frequencies of the blue and white morphs, in contrast, and 1993, whereas whites produced signi®cantly more seeds displayed much less temporal variation. At Pearblossom 1, than blues in 1991, 1995, and 1998 (Table 3; Fig. 1). After including only the seven years with a signi®cant density of the Bonferroni correction, three of the comparisons of mean plants, the frequency of the blue morph varied from 9.0 Ϯ seed number at Pearblossom 1 remain signi®cant (1988, 0.5% (SD) in 1998 to 15.9 Ϯ 1.2% in 1991 (Table 1). At 1993, and 1995; Table 3). White-¯owered plants at Pear- LINANTHUS FLOWER COLOR 1275

TABLE 2. Flower production by blue- and white-¯owered plants of Linanthus parryae in the Mojave Desert. The periods of observation at each site were: Pearblossom 1, 1988±1998; Pearblossom 2, 1991±1994; Antelope Valley, 1990±1993 (Table 1). Within these periods, plants were not sampled in years of low density (Table 1). Probability levels in bold face identify comparisons that are signi®cant after sequential Bonferroni tests conducted for each site (tablewide ␣-level of 0.05).

Number of ¯owers per plant Blue White Statistical comparisons of the two morphs1 Site Year Overall mean2 Mean SD Mean SD ZN3 P Pearblossom 1 1988 2.08 2.53 1.67 2.01 1.37 5.62 430 Ͻ0.0001 1989 1.23 1.28 0.55 1.22 0.50 1.05 100 0.29 1991 14.97 14.23 10.78 15.11 10.79 1.12 322 0.26 1992 2.96 3.15 2.09 2.93 2.08 1.21 138 0.23 1993 1.28 1.40 0.76 1.27 0.75 1.12 62 0.26 1995 4.38 4.41 2.93 4.37 2.67 0.03 416 0.98 1998 7.93 7.06 4.37 8.01 4.91 3.02 334 0.0026 Pearblossom 2 1991 8.35 7.92 4.61 8.64 5.17 2.55 451 0.0107 1992 3.96 4.07 3.46 3.90 2.93 0.40 145 0.69 1993 1.66 1.69 1.31 1.65 0.97 0.60 68 0.55 Antelope Valley 1990 3.23 3.80 3.18 3.08 2.13 2.25 88 0.0244 1991 4.87 4.88 4.48 4.87 5.27 0.87 374 0.39 1992 4.53 5.03 4.87 4.48 3.41 1.35 175 0.18 1993 1.49 1.46 0.90 1.50 0.79 0.86 294 0.39 1 Wilcoxon paired one-sample tests. 2 Calculated by weighting mean ¯ower number of each morph by its frequency (Table 1). 3 Number of sample pairs: one blue-¯owered and one white-¯owered plant per pair. blossom 2 produced signi®cantly more seeds than blues in To determine if there was signi®cant temporal variation in 1991. In 1992 the difference in seed production at this site relative seed production, for each site-year combination we was marginally signi®cant, with whites producing more seeds counted the number of sample pairs in which blue-¯owered than blues (Table 2; Fig. 2A). The magnitude of these dif- plants produced more seeds than whites and in which white- ferences was often quite large (Figs. 1, 2), with coef®cients ¯owered plants produced more seeds than blues. We then of selection against the disadvantaged morph as high as 0.60 conducted a ␹2-analysis for each site to determine if the two in some years (Table 3). By contrast, at Antelope Valley there categories of relative seed production (blue Ͼ white, white were no signi®cant differences in seed number between Ͼ blue) were independent of year. At Pearblossom 1, analysis morphs (Table 3; Fig. 2B). Fisher's combined probability of the 1880 sample pairs (ties not included) censused over test revealed a signi®cant overall effect of ¯ower color on seven years revealed that the relative seed production of blue- seed number at Pearblossom 1 (␹2 ϭ 62.2, df ϭ 14, P Ͻ and white-¯owered plants varied signi®cantly across years 0.0001) and at Pearblossom 2 (␹2 ϭ 25.3, df ϭ 6, P Ͻ 0.001), (␹2 ϭ 41.5, df ϭ 6, P Ͻ0.0001). In contrast, there was no but not at Antelope Valley (␹2 ϭ 4.9, df ϭ 8, P ϭ 0.77). signi®cant temporal variation in relative seed production at

TABLE 3. Seed production by blue- and white-¯owered plants of Linanthus parryae in the Mojave Desert. Sampling periods are explained in Table 2. Selection coef®cients, given for years/sites where there was a signi®cant difference in seed production between morphs, are calculated as 1 Ϫ (w1/w2), where w1 is the disadvantaged morph and w2 is the morph having higher seed production. Probability levels in bold face identify comparisons that are signi®cant after sequential Bonferroni tests conducted for each site (tablewide ␣-level of 0.05).

Number of seeds per plant Statistical comparisons of Blue White Selection the two morphs1 coeff. Site Year Overall mean2 Mean SD Mean SD (s) ZN3 P Pearblossom 1 1988 21.12 30.45 38.17 19.67 26.69 0.35 4.74 430 Ͻ0.0001 1989 3.99 5.00 8.99 3.83 7.23 0.59 131 0.550 1991 295.71 260.13 235.91 302.42 259.88 0.14 2.29 322 0.022 1992 14.06 18.58 40.26 13.28 35.49 0.29 2.12 359 0.034 1993 1.29 2.77 11.80 1.12 6.04 0.60 2.70 270 0.007 1995 72.46 65.62 72.88 73.35 73.69 0.11 2.74 463 0.006 1998 124.29 108.87 100.12 125.81 123.97 0.14 2.37 465 0.018 Pearblossom 2 1991 225.20 198.17 156.50 243.06 195.32 0.18 4.19 451 Ͻ0.0001 1992 19.32 16.58 38.72 20.86 39.10 0.21 1.85 406 0.064 1993 4.54 4.29 16.41 4.65 14.05 0.67 262 0.500 Antelope Valley 1990 38.11 43.18 62.54 36.74 45.42 0.04 88 0.970 1991 91.77 88.79 116.32 92.34 121.36 0.30 374 0.760 1992 45.59 49.45 67.11 45.21 68.41 1.14 244 0.250 1993 9.07 8.99 17.99 9.09 15.58 0.70 521 0.480 1 Wilcoxon paired one-sample tests. 2 Calculated by weighting mean seed number of each morph by its frequency (Table 1). 3 Number of sample pairs: one blue-¯owered and one white-¯owered plant per pair. 1276 D. W. SCHEMSKE AND P. BIERZYCHUDEK

FIG. 1. Mean seeds per plant for white-¯owered and blue-¯owered Linanthus parryae at Pearblossom 1 over the years 1988±1998. Means, standard deviations, and sample sizes are given in Table 3. NS, P Ͼ 0.05; *P Ͻ 0.05; **P Ͻ 0.01; ****P Ͻ 0.0001.

Pearblossom 2 (␹2 ϭ 3.27, df ϭ 3, P ϭ 0.20) or at Antelope 2B). For example, at Pearblossom 1 there was a 200-fold Valley (␹2 ϭ 3.00, df ϭ 4, P ϭ 0.39). range of variation in seed production over the seven years There was considerable variation in mean seed production with ¯owering plants, from a mean of 1.3 seeds/plant in 1993 per plant at all sites, as calculated by weighting each morph's to nearly 300 seeds/plant in 1991. These differences prompt- mean seed production by its frequency (Table 3; Figs. 1, 2A, ed us to ask if there was a relationship between the relative ®tness of the two morphs (mean seeds per plant for blue- ¯owered plants/mean seeds per plant for white-¯owered plants) and the quality of the year, as measured by the mean seed production of plants that year. We found signi®cant negative correlations between relative ®tness and mean seed production for all sites and years combined (rs ϭϪ0.62, N ϭ 14, P Ͻ 0.05; Fig. 3), for the combined data from the two Pearblossom sites (rs ϭϪ0.66, N ϭ 10, P Ͻ 0.05), and for Pearblossom 1 alone (rs ϭϪ0.86, N ϭ 7, P Ͻ 0.05). These results suggest that variation in precipitation among years might explain temporal variation in relative ®tness of the two morphs.

FIG. 2. Mean seeds per plant for white-¯owered and blue-¯owered Linanthus parryae. Means, standard deviations, and sample sizes FIG. 3. Mean relative ®tness (mean seeds per plant for blue-¯ow- are given in Table 3. NS, P Ͼ 0.05; ****P Ͻ 0.0001. Signi®cance ered plants/mean seeds per plant for white-¯owered plants) as a levels for marginally signi®cant comparisons are given above the function of mean seed number per ¯owering plant. Data from the bars. (A) Pearblossom 2 over the years 1990±1993. (B) Antelope same sites and years as Figures 1, 2A, B. The dashed line represents Valley over the years 1990±1993. equal ®tness for both morphs. LINANTHUS FLOWER COLOR 1277

Association between Plant Density, Fitness, and Mechanisms of Selection Precipitation On average, white-¯owered plants had the advantage in Both in our study and that of Epling et al. (1960), there good years, whereas blue-¯owered plants had higher ®tness were some years when ¯owering plants were absent or rare. in poor years. The mechanisms responsible for this temporal In these years, winter precipitation levels were probably too variation in the relative ®tness of the two morphs could op- low to trigger germination. In the greenhouse, Linanthus erate at a variety of life stages. For example, the morphs seeds could only be induced to germinate by simulating pro- might vary in their ability to produce large numbers of ¯ow- tracted periods of rainfall; simply keeping seeds moist was ers, in the percentage of their ¯owers that mature and produce insuf®cient (P. Bierzychudek, pers. obs.). Plant density in seeds, or in their seed production per ¯ower. These different our census transects was most highly correlated with precip- possibilities suggest different likely mechanisms of selection. itation in January (for all sites and years combined, rs ϭ 0.66, We ®rst investigated the relationship between seed pro- N ϭ 19, P ϭ 0.006), typically the month when seeds ger- duction and ¯ower production of the two morphs and found minate in the ®eld. We observed the highest plant density at that variation among sites and years in mean seeds per plant Pearblossom 1 in 1993, which had the wettest January in the was due largely to variation in mean ¯ower number. The last half century (Fig. 4A), with 13.6 times more precipitation correlation between mean seed number per plant and mean (19.1 cm) than the mean for the four years with few or no ¯ower number per plant was signi®cant for all sites combined plants (xÅ ϭ 1.4 cm). (rs ϭ 0.95, N ϭ 14, P Ͻ 0.001), for the combined data for To investigate the possibility that temporal variation in the two Pearblossom sites (rs ϭ 0.95, N ϭ 10, P Ͻ 0.01), precipitation is a cause of the signi®cant negative relationship and for Pearblossom 1 (rs ϭ 0.93, N ϭ 7, P Ͻ 0.05). We we observed between absolute and relative seed production also found a signi®cant, positive correlation (P Ͻ 0.0001) (Fig. 3), we sought to identify the precipitation pattern most between ¯ower and seed number per plant for each of the 14 highly correlated with plant ®tness. We found that the com- site-year combinations. bined precipitation in March and April gave the highest cor- In three of the four site-year combinations with a signif- relation with mean seed production for all sites combined (rs icant difference in ¯ower production between the two morphs ϭ 0.80, N ϭ 14, P ϭ 0.004; Fig. 5A), for the two Pearblossom (Table 2) we also observed a signi®cant difference in seed sites (rs ϭ 0.88, N ϭ 10, P ϭ 0.009), and for Pearblossom production (Table 3). For 10 of the 14 site-year combinations, 1(rs ϭ 0.86, N ϭ 7, P ϭ 0.036). March±April rainfall was the morph with the highest ¯ower production had the highest also signi®cantly correlated with relative ®tness (blue seed seed production. We also found a signi®cant positive cor- production/white seed production) for all sites combined (rs relation between the relative ®tness of the two morphs and ϭϪ0.60, N ϭ 14, P ϭ 0.032; Fig. 5B) and for the combined their relative ¯ower production (mean ¯owers per plant for data for the two Pearblossom sites (rs ϭϪ0.69, N ϭ 10, P blue-¯owered plants/mean ¯owers per plant for white-¯ow- ϭ 0.038). With increasing rainfall, the seed production of ered plants) for all sites and years combined (rs ϭ 0.75, N the blue morph decreased relative to that of the white morph. ϭ 14, P Ͻ 0.01), for the combined data from the two Pear- Even for Pearblossom 1 alone, the correlation between rel- blossom sites (rs ϭ 0.77, N ϭ 10, P Ͻ 0.01), and for Pear- ative ®tness and March±April rainfall was negative and mar- blossom 1 (rs ϭ 0.93, N ϭ 7, P Ͻ 0.05). These results suggest ginally signi®cant (rs ϭϪ0.71, N ϭ 7, P ϭ 0.080). that the differences in seed production between color morphs The precipitation patterns we observed during our study are due largely to mechanisms that in¯uence ¯ower produc- were comparable to longer-term trends (Figs. 4A, B). For the tion per plant, rather than to factors in¯uencing seed pro- period 1941±1987, the median January precipitation at Palm- duction per ¯ower. dale was 2.49 cm (mean ϭ 3.76 cm, SD ϭ 4.1 cm, range 0.00±15.9 cm), whereas the median for the period 1988 to Pollination 1998 was 2.79 cm (mean ϭ 4.91 cm, SD ϭ 5.9 cm, range 0.7±19.1 cm; Fig. 4A). These differences were not statisti- The beetle Trichochorous sp. (Melyridae) was the only cally signi®cant (Mann-Whitney U-test, Z ϭ 0.73, df ϭ 56, ¯ower visitor. We observed 131 pollinator visits in 1991 and P ϭ 0.463). Plants were rare or absent in four of the 11 years 271 visits in 1992. In each year pollinators exhibited slightly of study at Pearblossom 1 (Table 1; Fig. 4A), a frequency higher visitation to white-¯owered plants (55.7% of visits in (36%) nearly identical to that observed by Epling and col- 1991; 51.7% of visits in 1992), but this was not signi®cantly leagues (4/13 ϭ 31%; Epling and Dobzhansky 1942; Epling different from the null expectation in either year (1991: G et al. 1960). For the period 1941±1987, the median March± ϭ 1.72, df ϭ 1, P ϭ 0.20; 1992: G ϭ 0.30, df ϭ 1, P ϭ April precipitation at Palmdale was 3.86 cm (mean ϭ 4.79 0.59). cm, SD ϭ 4.34 cm, range 0.0±18.6 cm), and for 1988±1998 the median March±April precipitation was 2.52 cm (mean ϭ Water-use ef®ciency 3.79 cm, SD ϭ 3.50 cm, range 0.2±10.6 cm; Fig. 4B). Again, these differences were not statistically signi®cant (Mann- We found no signi®cant differences between morphs in Whitney U-test, Z ϭ 0.63, df ϭ 55, P ϭ 0.531). March±April their carbon isotope ratios. The mean WUE of the pooled precipitation patterns during the two time periods are also blue-¯owered samples was Ϫ27.24 Ϯ 0.32%, whereas the similar when we exclude those years in which January pre- mean of the white-¯owered samples was Ϫ27.13 Ϯ 0.30% cipitation was so low that germination was unlikely. (paired t-test, n ϭ 18, P ϭ 0.19). 1278 D. W. SCHEMSKE AND P. BIERZYCHUDEK

FIG. 4. Precipitation patterns for Pearblossom, California, during the course of the present study. (A) January precipitation levels in centimeters for each year of the present study. Years when ¯owering plants were absent or rare at Pearblossom 1 are indicated by text above the bars. Shown at the far right are the median and range for the 11 years of the present study (1988±1998) and for 1941±1987, a period that begins with the year of Epling and Dobzhansky's (1942) ®eld study and ends in the year preceding our study. (B) Combined March±April precipitation levels in centimeters for each year of the present study. Shown at the far right are the median and range for these 11 years and the median and range for the period 1941±1987 (see above). A comparison of the mean seed production by the two morphs at Pearblossom 1 (b, blue; w, white) is summarized for each year with a signi®cant density of ¯owering plants (data from Table 3). Years when one morph produced signi®cantly more seeds per plant than the other are indicated by text above the bars.

DISCUSSION seed bank, exclusively outcrossing mating system, and high ¯owering plant densities (up to 67 plants/m2 at Pearblossom Sewall Wright believed that ¯ower color in L. parryae was 1; Table 1) are not characteristics typically associated with a nearly neutral character (Wright 1978). He found evidence low effective population size (Crawford 1984). of only slight selective differences between the morphs (s Յ 0.01), and these were manifest only at the largest spatial Spatial and Temporal Variability in Selection scales. The effective population size that he estimated for this species was small enough for random genetic drift to At Pearblossom 1, where from 1988 to 1998 we observed override such a small selective difference between the morphs seven Linanthus ¯owering seasons, blue-¯owered plants pro- (Wright 1978). He proposed that the combination of weak duced signi®cantly more seeds than white-¯owered plants in selection and small effective population size in L. parryae three years, and white-¯owered plants outproduced blue would result in isolation by distance, that is, the loss of ge- plants in three years. At Pearblossom 2, we found that white- netic variation within populations and the evolution of ge- ¯owered plants produced signi®cantly more seeds than blue- netic differentiation among populations. The spatial pattern ¯owered plants in two of three years. At Antelope Valley, of ¯ower color observed in L. parryae was thought to conform we never found a signi®cant difference between morphs in to these predictions (Provine 1986). seed production. Thus, only at Antelope Valley are we unable In contrast, our data demonstrate that ¯ower color in L. to reject Wright's contention that ¯ower color is neutral and parryae is sometimes subject to strong selection. At Pear- subject only to the stochastic process of genetic drift. blossom 1, the average selection coef®cient (s) was 0.23 (in- The L. parryae population at Antelope Valley behaved cluding all years with ¯owering plants; Table 3). Although somewhat differently from those at Pearblossom, ¯uctuating it is theoretically possible for genetic drift to override selec- less in density from year to year. In 1990, when no plants tion of this magnitude, it would require an extremely small were observed at Pearblossom 1, plant density at Antelope effective population size. The conditions for genetic drift to Valley was reasonably high (Table 1). Plants at Antelope override selection, Nes Ͻ 1, are satis®ed at Pearblossom 1 Valley were smaller and produced fewer ¯owers and seeds, only if the effective population size (Ne) is less than 4.4. We and differences in seed production between the two morphs do not have an estimate of the effective population size in were not as great as at Pearblossom. We also observed that L. parryae, so we cannot determine precisely if genetic drift the intensity of color in blue ¯owers at Antelope Valley was could override the selection observed in our studies. Nev- more variable than at Pearblossom, where the blue and white ertheless, as suggested by Epling et al. (1960), it seems very morphs were strikingly different. These subtler differences unlikely that the effective population size of L. parryae is in color were associated with correspondingly smaller dif- small enough to negate the effects of selection. Its long-lived ferences in performance at Antelope Valley. LINANTHUS FLOWER COLOR 1279

and may represent a major cause of the temporal variation in selection we witnessed at Pearblossom 1. The 1993 El NinÄo produced high January precipitation but low March± April precipitation at Pearblossom 1, thus the conditions in this year favored blue-¯owered plants (Figs. 4A, B). In con- trast, each of the remaining El NinÄo years produced high March±April precipitation, and these were also the only years of white advantage at Pearblossom 1 (1991, 1995, and 1998; Fig. 1).

Mechanisms of Selection Levels of spring rainfall appear to have very different ef- fects on the ®tness of the two color morphs in our study populations. In particular, at Pearblossom 1 the relative seed production of blue-¯owered plants was highest in drier years, whereas that of white-¯owered plants was highest in wetter years. This suggests an interaction between ¯ower color ge- notype and environment. Pollinators could cause selection on ¯ower color if they visited one color morph more frequently than another, yet we found no signi®cant color preference in either of our two years of pollinator observations. We measured pollinator vis- itation to the two morphs in 1991, a year of very high seed production and white advantage at Pearblossom 1, and in 1992, a year in which this site displayed much lower seed production and blue advantage. Pollinators in both years vis- ited the two morphs at nearly equal rates; therefore it does not appear that the temporal variation in selection that we observed at Pearblossom 1 can be explained by changes in pollinator preference. Because the frequencies of the two morphs did not change appreciably between the two years in which we observed pollinators (1991, 1992; Table 1), we cannot rule out the possibility that pollinators might dem- onstrate frequency-dependent color preferences, a behavior that could contribute to the maintenance of the polymor- phism. We suggest that pollinators probably do not contribute FIG. 5. (A) Mean seeds per plant, with both color morphs pooled, to selection on ¯ower color in L. parryae, but caution that plotted as a function of combined March and April rainfall. (B) more detailed study is warranted. For example, it would be Relative ®tness (mean seeds per plant for blue-¯owered plants/mean of interest to determine if pollinator foraging behavior differs seeds per plant for white-¯owered plants) plotted as a function of combined March and April rainfall. The dashed line represents equal between the color morphs, due perhaps to morph-speci®c ®tness for both morphs. variation in pollen production, and if this could cause dif- ferences between morphs in their male or female reproductive success. For all populations taken together, we observed a negative Factors in¯uencing plant growth are the most likely ex- correlation between relative ®tness and mean seed production planations for the observed temporal differences in plant size. of the morphs and a signi®cant positive correlation between We found no evidence that the two morphs differ in WUE, mean seed production and March±April precipitation. Thus, as indicated by their carbon isotope ratios. The substantial the performance of blue- and white-¯owered plants appears cost of these analyses required that we limit our investigation to be determined by some temporally varying environmental to only a single year (1998) in just one site (Pearblossom 1). factor, most likely precipitation. This is perhaps best illus- This was a year of high spring precipitation, so it would be trated by the seed production data from Pearblossom 1, where useful to have a comparative sample from a drier year. Be- the mean March±April precipitation for the three years of cause of the limited sampling and lack of replication across white advantage was nearly twice that for the years of blue sites and years, the ®nding of no difference between the advantage (7.96 cm vs. 4.50 cm; Fig. 4B). Four years of our morphs in WUE should be viewed with caution. study (1991, 1993, 1995, and 1998) were classi®ed as El There are countless other possible pleiotropic effects of NinÄo, the tropical climatic phenomenon that often results in ¯ower color that may explain the ®tness differences of the a marked increase in winter precipitation along the western two morphs. Recent ®ndings suggest that white-¯owered Paci®c (Trenberth 1997; N. Mantua, pers. comm.). El NinÄo plants possess a higher tissue concentration of potentially occurs in this region every three to 12 years (Quinn 1987), toxic cations such as magnesium and selenium and that pop- 1280 D. W. SCHEMSKE AND P. BIERZYCHUDEK ulation differentiation for ¯ower color is associated with spa- (Rausher et al. 1993), might also contribute to the mainte- tial variation in soil chemistry (D. W. Schemske, unpubl. nance of the polymorphism. A complete discussion of the data). These surprising results suggest a hypothesis to explain mechanisms contributing to the maintenance of the ¯ower- the relationship between relative ®tness and precipitation that color polymorphism at Pearblossom 1 is given in Turelli et we observed at Pearblossom 1 (Fig. 1). In years with low al. (2001). soil moisture, the tissue concentrations of detrimental cations The morph frequency in the area surrounding Pearblossom in white-¯owered plants may become high enough to reduce 1 is very similar to that observed in our transects at this site, their ®tness, whereas blue-¯owered plants are less adversely so it is unlikely that the ¯ower color polymorphism there is affected, by virtue of their reduced cation uptake. In years due to gene ¯ow from adjacent populations. This is not the with high soil moisture, detrimental cations become diluted, case for Pearblossom 2, which has neighboring populations and blue-¯owered plants may display lower ®tness because that are primarily blue- and primarily white-¯owered. Here, their mechanism of cation exclusion also reduces the uptake with only three years of data on seed production, we have of favorable minerals such as calcium and phosphorus. A not observed ¯uctuating selection. Instead, we observed similar mechanism could explain the marked spatial variation strong selection against white-¯owered plants in one year in ¯ower color observed by Epling and Dobzhansky (1942) and a marginally signi®cant advantage for this morph in an- and Epling et al. (1960). Experimental studies are now un- other year. If selection over longer time periods typically derway to test this hypothesis in L. parryae and in other favors white-¯owered plants, the ¯ower color polymorphism congeners that display a similar ¯ower-color polymorphism. at Pearblossom 2 could be maintained by gene ¯ow from neighboring blue-¯owered populations. Maintenance of the Polymorphism At Antelope Valley, in four years of observation there was never a signi®cant advantage for one ¯ower color morph over Wright believed that random genetic drift was the principle the other. However, ¯ower color at this site was more var- mode of evolution within L. parryae populations and that this iable, suggesting a possible role for modi®er genes. At this process was a suf®cient explanation for the spatial distri- site we observed ¯owering in years where there were no bution of ¯ower color. The geographical surveys conducted plants at the Pearblossom sites, and a range of seed produc- by Epling and Dobzhansky (1942) revealed that polymorphic tion that was lower than that at Pearblossom. Although we populations of L. parryae were rather rare. According to this cannot reject the hypothesis that ¯ower color at Antelope argument, polymorphic populations should quickly become Valley is a neutral character, further study is clearly war- ®xed for alternate alleles, and this would result in spatial ranted. differentiation for ¯ower color. One possible alternative to this explanation is that tem- The Historical Debate porally varying natural selection is actively maintaining poly- morphic populations. The requirements for ¯uctuating selec- Epling and Dobzhansky (1942) ®rst concluded that the tion to maintain are rather restrictive (Hal- spatial pattern of ¯ower color in L. parryae was consistent dane and Jayakar 1963; Hedrick et al. 1976; Hedrick 1986), with Wright's prediction that random genetic drift could but they become signi®cantly less so with overlapping gen- cause the genetic differentiation of populations. Wright erations (Templeton and Levin 1979; Ellner and Hairston (1943) followed their paper with a detailed reanalysis of their 1994; Ellner and Sasaki 1996). The long-lived seed bank of data, concluding that L. parryae was an example of isolation L. parryae provides such age structure. by distance, the initial step in the shifting balance process. For Pearblossom 1, where we have the longest record of However, there was an early suggestion that this conclusion plant performance, there is clear evidence that selection is might be incorrect. William Hovanitz, a graduate student at both strong and temporally variable. Recent theoretical anal- Cal Tech who had worked with Dobzhansky on Drosophila, ysis of these data (Turelli et al. 2001) suggests that ¯uctuating wrote to Wright in 1942: ``it seems to me very dif®cult to selection, coupled with a long-lived seed bank, can maintain imagine a gene existing in wild populations which has ab- the polymorphism. Speci®cally, both morphs are maintained solutely no different physiologic effect on individuals car- if the genotype whose arithmetic and geometric mean relative rying it as compared with a type standardized as `wildtype.''' ®tness are both less than one also has a relative ®tness greater Hovanitz went on to describe what he knew of L. parryae than one in years with its highest seed production. This is populations in the ®eld and concluded: ``there is fairly good precisely what we observed at Pearblossom 1, where white- reason to suppose, therefore, that the blue ¯owered plants ¯owered plants have both the lowest arithmetic and geometric are being selected for in these areas'' (Provine 1986, p. 375). mean relative ®tness, but are maintained in the population After nearly 20 years of additional ®eld studies, this was because of their ®tness advantage in the years of highest seed precisely the conclusion reached by Epling and his colleagues production (Turelli et al. 2001). Precipitation patterns during (Epling et al. 1960). They suggested that the earlier conclu- our 11 years of observation at Pearblossom 1 were compa- sions reached by Epling and Dobzhansky (1942) and Wright rable to those over the last half century, thus the ¯uctuating (1943) were wrong, and proposed the alternative explanation selection we documented at this site may be suf®cient to that the ¯ower color polymorphism in L. parryae was the maintain the polymorphism. Nevertheless, we cannot rule out product of natural selection (Epling et al. 1960). As described the possibility that other mechanisms, such as heterozygote by Provine (1986), Wright was dismayed by these new ®nd- superiority due to pleiotropy (Rausher and Fry 1993) or fre- ings, but declined Epling's invitation to contribute a technical quency-dependent selection mediated by pollinator behavior appendix to his paper. Wright reanalyzed the entire dataset LINANTHUS FLOWER COLOR 1281 and published the results many years later, staunchly main- El NinÄo; Pomona College students E. Chen, A. Scherer Des- taining his view that natural selection on ¯ower color was marais, H. Foshee, B. Godsey, J. Astle Happy, S. Im, H. of only minor signi®cance in L. parryae (Wright 1978). Merchant, K. Merriam, and Y. Smith together counted hun- One possible explanation for Wright's persistence is that dreds of thousands of Linanthus seeds; dozens more Pomona Epling et al. (1960) failed to conduct detailed, direct obser- College students provided ®eld and/or laboratory assistance; vations, and instead focused more on the possible causes of D. Boose, P. Pack, and B. Best helped with many aspects of selection than on selection itself. For example, they inves- the ®eldwork; R. Levin provided us with a base camp in tigated whether there were differences in soils between all- southern California; and The Pines provided an oasis of sanity blue and all-white areas and found none. Likewise, they com- in the biological and cultural desert of Palmdale. We thank pared the composition of the plant community in all-blue and J. Coyne, H. Lewis, M. Turelli, and R. Lande for useful all-white areas and again found no differences. Their con- discussion throughout the course of our studies; P. Miksova clusion that ¯ower color in L. parryae is subject to selection and N. R. Temkin for statistical advice; and two anonymous was based on experimental evidence that this species has a reviewers for helpful comments on the manuscript. large, long-lived seed bank and on long-term census data Financial support was provided by National Science Foun- showing that morph frequencies were temporally stable. Even dation BSR-8918246 to the University of Washington (DWS) in their own minds, this was not strong evidence for selection: and National Science Foundation RUI BSR-8919674 to Po- ``the frequencies of blue and white ¯owered plants are in the mona College (PB). Other support was provided by Pomona long run the product of selection operating at an intensity we College Schenck Funds and a Pomona College Seaver Grant have been unable to measure'' (Epling et al. 1960, p. 254). to PB. DWS and PB thank their ancestors for the stress- Wright had another reason for doubting the conclusions of tolerance traits that allowed them to tolerate the desert's Epling et al. (1960). Linanthus parryae was Wright's ``®rst prickliness as well as each other's. and best example of isolation by distance'' (Provine 1986, p. 485) and ``an example of the whole shifting balance theory LITERATURE CITED of evolution in nature in action'' (Provine 1986, p. 378). With Barton, N. H. 1992. On the spread of new combinations in the third so much at stake, Wright was apparently unable to consider phase of Wright's shifting balance. Evolution 46:551±557. objectively the conclusions of Epling et al. (1960) that ¯ower Brown, B. A., and M. T. Clegg. 1984. The in¯uence of ¯ower color color in L. parryae was not the result of isolation by distance. polymorphisms on genetic transmission in a natural population of the common morning glory, Ipomoea purpurea. Evolution 38: It is hard to fathom how Wright could give so little merit to 796±803. Epling's ®ndings. Wright clearly valued the census data col- Burdon, J. J., D. R. Marshall, and A. H. D. Brown. 1983. Demo- lected by Epling and his colleagues, as it formed the basis graphic and genetic changes in populations of Echium planta- for all his own analyses, yet he apparently saw no inconsis- ginneum. J. Ecol. 71:667±679. tency in ignoring virtually all of their other data. Coyne, J. A., N. H. Barton, and M. Turelli. 1997. Perspective: a critique of Sewall Wright's shifting balance theory of evolution. Despite considerable controversy, the shifting balance the- Evolution 51:643±671. ory remains one of the most celebrated of Wright's contri- Crawford, T. J. 1984. What is a population? Pp. 135±173 in B. butions to evolutionary biology (Provine 1986; Crow 1991; Shorocks, ed. Evolutionary ecology. Blackwell Scienti®c, Ox- Wade and Goodnight 1991; Coyne et al. 1997). His detailed ford, U.K. Crow, J. 1991. Was Wright right? Science 253:973. analysis of the ¯ower color polymorphism in L. parryae was Crow, J., W. R. Engels, and C. Denniston. 1990. Phase three of pivotal to the development of this theory (Provine 1986). We Wright's shifting balance theory. Evolution 44:233±247. provide strong evidence that Wright greatly underestimated Ehleringer, J. R. 1988. Carbon isotope ratios and physiological the role of natural selection in polymorphic populations of processes in aridland plants. Pp. 41±54 in: P. W. Rundel, J. R. L. parryae. This conclusion is supported by additional studies Ehleringer, and K. A. Nagy, eds. Stable isotopes in ecological research. Springer, New York. showing that a sharp, local discontinuity in ¯ower color is Ehleringer, J. R., J. P. Comstock, and T. A. Cooper. 1987. Leaf- associated with a number of environmental differences and twig carbon isotope ratio differences in photosynthetic-twig de- by transplant studies that demonstrate selection against the sert shrubs. Oecologia 71:318±320. rare color morph (D. W. Schemske and P. Bierzychudek, Ellner, S., and N. G. Hairston Jr. 1994. Role of overlapping gen- erations in maintaining variation in a ¯uctuating environment. unpubl. ms.). Our results for L. parryae are therefore con- Am. Nat. 143:403±441. sistent with Mayr's suggestion that ``Selective neutrality can Ellner, S., and A. Sasaki. 1996. Patterns of genetic polymorphism be excluded almost automatically wherever polymorphism or maintained by ¯uctuating selection with overlapping genera- character clines are found in natural populations'' (Mayr tions. Theor. Popul. Biol. 50:31±36. 1965, p. 207). Endler, J. A. 1986. Natural selection in the wild. Princeton Univ. Press, Princeton, NJ. Epling, C., and T. Dobzhansky. 1942. Genetics of natural popu- ACKNOWLEDGMENTS lations. IV. Microgeographic races in Linanthus parryae. Ge- netics 27:317±332. Over the 11 years of this project, we received assistance Epling, C., H. Lewis, and F. M. Ball. 1960. The breeding group from many individuals and institutions, only some of which and seed storage: a study in population dynamics. Evolution 14: we are able to mention here. H. Lewis showed us the location 238±255. of the original study sites; C. Taylor allowed us to see C. Ernst, W. H. O. 1987. Scarcity of ¯ower colour polymorphism in ®eld populations of Digitalis purpurea L. Flora 179:231±239. Epling's maps and ®eld notes; S. Curtis and E. Soulanille Farquhar, G. D., K. T. Hubick, A. G. Condon, and R. A. Richards. provided expert ®eld and laboratory assistance for several 1988. Carbon isotope fracionation and plant water-use ef®cien- key years of the study; N. Mantua provided information on cy. Pp. 21±40 in: P. W. Rundel, J. R. Ehleringer, and K. A. 1282 D. W. SCHEMSKE AND P. BIERZYCHUDEK

Nagy, eds. Stable isotopes in ecological research. Springer, New W. Fairbridge, eds. The encyclopedia of climatology. Van Nos- York. trand Reinhold, New York. Fisher, R. A. 1922. On the dominance ratio. Proc. R. Soc. Edin- Rausher, M. D., and J. D. Fry. 1993. Effects of a locus affecting burgh. 42:321±341. ¯oral pigmentation in Ipomopsis purpurea on female ®tness com- ÐÐÐ. 1930. The genetical theory of natural selection. 2d ed. ponents. Genetics 134:1237±1247. Oxford Univ. Press, Oxford, U.K. Rausher, M. D., D. Augustine, and A. VanderKool. 1993. Absence ÐÐÐ. 1954. Statistical methods for research workers. 12th ed. of pollen discounting in genotypes of Ipomoea purpurea exhib- Oliver and Boyd, Edinburgh. iting increased sel®ng. Evolution 47:1688±1695. Fisher, R. A., and E. B. Ford. 1947. The spread of a gene in natural Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution conditions in a colony of the moth Panaxia dominula. Heredity 43:223±225. 1:143±174. Schoen, D. J., D. E. Giannasi, R. A. Ennos, and M. T. Clegg. 1984. Goodnight, C. J. 1995. Epistasis and the increase in additive genetic Stem color and pleiotropy of genes determining ¯ower color in variance: implications for phase 1 of Wright's shifting-balance the common morning glory. J. Hered. 75:113±116. process. Evolution 49:502±511. Snedecor, G. W., and W. G. Cochran. 1967. Statistical methods. Grant, V. 1981. Plant speciation. 2d ed. Columbia Univ. Press, New 6th ed. Iowa State Univ. Press, Ames. York. Sokal, R. R., and F. J. Rohlf. 1994. Biometry. 3rd ed. W. H. Freeman Haldane, J. B. S. 1932. The causes of evolution. Longman, London. and Co., San Francisco, CA. Haldane, J. B. S., and S. D. Jayakar. 1963. Polymorphism due to Stanton, M. L. 1987. Reproductive biology of petal color variants in wild populations of Raphanus sativus. I. Pollinator response selection of varying direction. J. Genetics 58:237±242. to color morphs. Am. J. Bot. 74:178±187. Hedrick, P. W. 1986. Genetic polymorphism in heterogeneous en- Stanton, M. L., A. A. Snow, S. N. Handel, and J. Bereczky. 1989. vironments: a decade later. Annu. Rev. Ecol. Syst. 7:535±566. The impact of a ¯ower-color polymorphism on mating patterns Hedrick, P. W., M. E. Ginevan, and E. P. Ewing. 1976. Genetic in experimental populations of wild radish (Raphanus raphan- polymorphism in heterogeneous environments. Annu. Rev. Ecol. istrum L.). Evolution 43:335±346. Syst. 7:1±32. Stebbins, G. L., Jr. 1950. Variation and evolution in plants. Co- Kay, Q. O. N. 1978. The role of preferential and assortative pol- lumbia Univ. Press, New York. lination in the maintenance of ¯ower colour polymorphisms. Pp. Taylor, L. P., and R. Jorgensen. 1992. Conditional male fertility in 175±190 in A. J. Richards, ed. Pollination and evolution. Royal chalcone synthase-de®cient Petunia. J. Hered. 83:11±17. Botanic Gardens, Sydney. Templeton, A. R., and D. A. Levin. 1979. Evolutionary conse- Kondrashov, A. S. 1992. The third phase of Wright's shifting bal- quences of seed pools. Am. Nat. 114:232±249. ance: a simple analysis of the extreme case. Evolution 46: Trenberth, K. E. 1997. The de®nition of El NinÄo. Bull. Am. Meteor. 1972±1975. Soc. 78:2771±2777. Levin, D. A. 1988. Local differentiation and the breeding structure Turelli, M., D. W. Schemske, and P. Bierzychudek. 2001. Stable of plant populations. Pp. 305±329 in L. D. Gottlieb and S. K. two-allele polymorphisms maintained by ¯uctuating ®tnesses Jain, eds. Plant evolutionary biology. Chapman and Hall, Lon- and seed banks: protecting the blues in Linanthus parryae. Evo- don. lution 55:1283±1298. Levin, D. A., and E. T. Brack. 1995. Natural selection against white Wade, M. J., and C. J. Goodnight. 1991. Wright's shifting balance petals in Phlox. Evolution 49:1017±1022. theory: an experimental study. Science 253:1015±1018. Martin, B., and Y. R. Thorstenson. 1988. Stable carbon isotope Waser, N. M., and M. V. Price. 1985. Reciprocal transplant ex- composition (␦13C), water use ef®ciency, and biomass produc- periments with Delphinium nelsonii (Ranunculaceae): evidence tivity of Lycopersicon esculentum, Lycopersicon pennellii, and for local adaptation. Am. J. Bot. 72:1726±1732. Wright, S. 1931. Evolution in Mendelian populations. Genetics 16: the F1 hybrid. Plant Physiol. 88:213±217. Mayr, E. 1942. Systematics and the origin of species. Columbia 97±159. Univ. Press, New York. ÐÐÐ. 1943a. Isolation by distance. Genetics 28:114±138. ÐÐÐ. 1965. Animal species and evolution. Belknap Press of Har- ÐÐÐ. 1943b. An analysis of local variability of ¯ower color in vard Univ. Press, Cambridge, MA. Linanthus parryae. Genetics 28:139±156. Patterson, R. 1993. Linanthus. Pp. 840±844 in J. C. Hickman, ed. ÐÐÐ. 1978. Evolution and the genetics of populations. VI. Var- The Jepson manual; higher plants of California. Univ. of Cali- iability within and among natural populations. Univ. of Chicago fornia Press, Berkeley. Press, Chicago. ÐÐÐ. 1982. The shifting balance theory and macroevolution. Provine, W. B. 1986. Sewall Wright and evolutionary biology. Univ. Annu. Rev. Genet. 16:1±19. of Chicago Press, Chicago. Quinn, W. H. 1987. El NinÄo. Pp. 411±414 in J. E. Oliver and R. Corresponding Editor: C. Eckert